Description

Aspire Software is looking for a data scientist to join our team in Lebanon.

Here is a little window into our company: Aspire Software operates and manages wholly owned software companies, providing mission-critical solutions across multiple verticals. By implementing industry best practices, Aspire delivers a time sensitive integration process, and the operation of a decentralized model has allowed it to become a hub for creating rapid growth by reinvesting in its portfolio.

About the job

We're looking for a data scientist who has taken machine learning models — especially reinforcement learning — from research to production. Today, our pricing engine is a rule-based parametric system (elasticity modeling, sigmoid demand curves, day-of-week weighting, occupancy and pickup deviation guardrails). Your job is to evolve it into a learning system: contextual bandits, RL policies, and probabilistic forecasting that price thousands of hotel-room-nights every day. You will also integrate other signals into this forecasted price, like competitor prices, events in the area, weather, etc.

You'll own this work end-to-end: framing the problem, designing rewards and offline evaluation, training models, and shipping them as production Python services on our FastAPI / AWS stack — not handing notebooks to engineers. You'll be expected to move fast using AI-assisted development tools.

What you'll work on

• Pricing intelligence — replace and extend our parametric pricing engine (occupancy deviation, pickup velocity, price elasticity, booking curve forecasting, seasonality, day-of-week effects) with learned models: contextual bandits, RL policies, and Bayesian elasticity estimation
• RL in production — design reward functions, exploration strategies, and off-policy evaluation that let us deploy RL pricing safely across multi-tenant hotel data; build the training, monitoring, and rollback infrastructure to support it
• Demand forecasting — improve our booking-curve and final-occupancy forecasts (currently sigmoid-based) with proper time-series and probabilistic methods; quantify uncertainty and feed it into pricing decisions
• Simulation & evaluation — extend our historical replay and synthetic simulation harness into a first-class offline evaluation and A/B testing framework for pricing policies
• LLM-powered features — build agentic workflows (OpenAI, Anthropic Claude, LangChain / LangGraph) for event-based pricing recommendations, demand analysis, and revenue-manager copilots
• Productionization — write production-grade Python services: typed, tested, modular packages running on FastAPI / SQLAlchemy / PostgreSQL — the kind of code a staff engineer would approve, not scripts and notebooks thrown over the wall
• Data pipelines — work with PredictHQ event data, competitor rate feeds, and PMS integrations (Seekda, InnQuest, others) to build reliable data flows that power pricing decisions
• Infrastructure — contribute to our AWS architecture (ECS Fargate, SQS, EventBridge, S3, CloudWatch) and help scale the platform as we grow

Tech stack

• Core: Python 3.11, FastAPI, SQLAlchemy 2.0, Alembic, PostgreSQL, Redis
• ML / RL: PyTorch or TensorFlow, scikit-learn, Stable-Baselines3 / Ray RLlib (or equivalent), MLflow or similar experiment tracking
• AI / LLM: OpenAI GPT-4, Anthropic Claude, LangChain, LangGraph, PredictHQ
• Data: Pandas, Polars, NumPy, statsmodels
• Infrastructure: AWS (ECS Fargate, SQS, EventBridge, S3, CloudWatch, ECR), Docker, GitHub Actions CI/CD
• Observability: Prometheus, Grafana Loki, PostHog

Requirements

• 4+ years of professional data science / ML engineering experience with models running in production (not just notebooks, dashboards, or analytics)
• Production reinforcement learning experience — you have personally designed, trained, deployed, and monitored at least one RL or contextual-bandit system serving real users at scale. You can speak in detail to: reward design, exploration / exploitation trade-offs, off-policy evaluation, distribution shift, safe rollout, and what broke when the model met production
• Strong Python development skills beyond scripting and Jupyter — you write modular, typed, tested Python packages; you're comfortable with async patterns, ORMs (SQLAlchemy), building production APIs (FastAPI or similar), and you can hold your own in a code review with backend engineers
• Solid foundations in classical ML, statistics, and time-series — regression, Bayesian methods, causal inference, demand forecasting, price elasticity
• Experience working with LLMs (OpenAI, Anthropic, or similar) and frameworks like LangChain or LangGraph for agentic workflows
• AI-assisted development is a must — you actively use tools like Claude Code, Cursor, GitHub Copilot, or similar to accelerate your workflow. We expect you to ship faster and think bigger because of these tools
• Strong SQL and data-modeling skills (PostgreSQL preferred)
• Experience with AWS cloud services or equivalent cloud platforms
• Comfortable working with Docker, CI/CD pipelines, and production deployments

Nice to have

• Experience in revenue management, hospitality tech, dynamic pricing, yield optimization, or ad / e-commerce bidding
• Background in price elasticity estimation, contextual bandits for pricing or recommendation, or hierarchical Bayesian demand models
• Experience with event-driven architectures (SQS, EventBridge, or similar)
• Familiarity with model and data observability — Prometheus / Grafana, drift detection, model performance dashboards
• Experience building multi-tenant SaaS platforms
• Publications, open-source contributions, or competition results in ML / RL

What we value

• Speed with quality — ship fast, but ship code and models a staff engineer would approve
• AI-native workflow — you don't just know about AI tools, you use them daily to write, debug, and architect
• Ownership — pick up a problem and drive it to completion without hand-holding
• Simplicity — elegant solutions over over-engineered ones. Minimal code that does the job
• Curiosity — our domain (hotel revenue optimization) has real depth. You're excited to learn it